[BioC] The reported bug in affy package persists

[Hongxian He]

Today I independently identified a bug in the pmcorrect.mas method from 
"affy" package (v1.1.1), which has been posted on the mailing list before. 
But I found that the bug is still in the lastest release of affy package, 
and the results computed therefore disagree with MAS5. 


pm.corrected <- apply(cbind(pps.pm, pps.im, delta), 1, max)
  {
	...
	 pm.corrected <- apply(cbind(pps.pm, pps.im, delta), 1, max)
	...
  }

Shouldn't the above line be 

	pm.corrected <- apply(cbind(pps.pm - pps.im, delta), 1, max) 

according to the formula V_{i,j} = max(PM_{i,j} - IM_{i,j}, d) in 
Affymetrix statistical algorithm document, where PV_{i,j}= log2(V_{i,j}), 
PV_{i,j} is the probe value for probe pair j in probeset i?

Thanks,

Hongxian He
Center of Bioinformatics
University of Pennsylvania


==================================================================
Quoted from the archive:

[BioC] the result of bioconductor not agree with MAS
"Hüsing, Johannes" johannes.huesing at medizin.uni-essen.de
Mon Mar 10 18:04:43 MET 2003

> -----Original Message-----
> From: Ben Bolstad [mailto:bolstad at stat.berkeley.edu]
> Sent: Thursday, March 06, 2003 5:48 PM
> To: Lang Chen
> Cc: 'bioconductor at stat.math.ethz.ch'
> Subject: Re: [BioC] the result of bioconductor not agree with MAS
> 
> 
> Hi Lang Chen (and to anyone else this is an issue for),
> 
> We have made our best attempts to duplicate the MAS 5.0 
> algorithm using
> available documentation. 

If I read the code in pmcorrect.mas() correctly (affy version 1.1), 
the mismatch is not subtracted from the perfect match, so the line

pm.corrected <- apply(cbind(pps.pm, pps.im, delta), 1, max)

should IMHO rather read

pm.corrected <- apply(cbind(pps.pm-pps.mm, pps.im-pps.mm, delta), 1,
max) ?

I might very well be missing something, so if somebody could give
me a clue I'd be grateful.

Best regards


Johannes